Evaluation of Fire Intensity Based on Neural Networks in a Forest-Fire Monitoring System

摘要

This work presents a method of fire intensity evaluation based on uncertain observations obtained by a forest fire monitoring system using jointly UAVs, remote sensing, and image processing. The proposed method uses two backpropagation neural networks to smoke and flame recognition. Flame color features, as well as dynamic shape/edge changing features, are trained using neural networks. Smoke and precipitation considered as the noise, various factors such as weather, light, features of sensors used in UAV were taken into account. The developed method provides sufficient performance to fulfill the requirements of credibility and efficiency.